Discover more from Karlstack
BREAKING: American Economic Review retracts Princeton economics professor's paper due to "coding error"
"The dog ate my code!" example #48584023
Following this Harvard bombshell from 2 days ago…
BREAKING: HBS professor placed on "administrative leave" following bombshell investigation into fake data
…today a new revelation has emerged from Princeton, which marks the fourth Princeton article I have covered.
Support independent journalism
Today, the focus is on Adrien Matray, an assistant professor at Princeton's Department of Economics, and a French citizen. While the paper in question, "Dividend Taxes and the Allocation of Capital", was co-authored with Charles Boissel, Matray is facing the heat all alone because Boissel has already departed from academia.
This paper has been retracted as of today.
Surprisingly, there has been no reporting on it yet, nor has it garnered any attention on Twitter. It will blow up soon, as soon as #EconTwitter catches wind of it. I only found out about this retraction because the story first broke on an obscure economics forum.
A few points:
“The errors described in this notice, which are responsible for the retraction, are visible in that code and were brought to the authors’ attention by Bach et al. (2023).”
A team of 4 French economists replicated this paper and found the fatal error.
“The firm size control variable in the conditionally accepted manuscript was coded in changes but was incorrectly written in the baseline specification as levels. This error was caught in the replication process when the data were rebuilt from scratch, and the variable was redescribed accordingly in the paper
A line of code was “incorrectly written”.
“Second, the code that plots Figure 4 in the publicly available code repository includes an alteration of two coefficients by a factor of 1.8, which was incorrect. This alteration affects the rendering of the figure, but not the underlying data. All the results reported in the tables and other figures are unaffected. The error in the rendering of the figure was not intentional.”
Another convenient coding error!
An anonymous economist says: “Kinda cringe. Dividing two arbitrarily selected coefficients by an arbitrary number 1.8 will never be considered unintentional … You divided exactly 2 coefficients by 1.8 and you know which two they were. Rarely seen a more obvious case of intentional fraud.”
"The error in the rendering of the figure was not intentional. The data and the code for the analysis are stored on remote, external servers that were inaccessible to the authors for prolonged periods of time during the publication process so it is unfortunately not possible to recover past versions to isolate the error’s origin."
LOL. The dog ate his code.
Based on this retraction I would say this dude is almost definitely guilty of data fraud, but, he has a modicum plausible deniability. There is a real chance this is all just due to incompetence, or a brainfart. I doubt it, but still.
There is not much new in this response, other than the fact that he admits the editor made him retract it, he didn’t want to retract it, rather he “believes that the scientific concerns could have been addressed in a reply.” The tone of the letter is really bitter, lol, he is clearly mad the editors didn’t sweep it under the rug for him.
Perhaps the worst part of the Twitter statement is that he makes it sound like this coding error has no consequences for the conclusions, which is a lie. His results were destroyed, and he is trying to pretend like they weren’t. He then goes on to explain his fortuitous coding errors, except he didn’t really explain anything, or apologize, he basically just said that life happens sometimes, okay? And we can’t look any further into it, because “the data and code for the analysis were stored on remote, external servers that were inaccessible”.
He then goes on to explain how part of his “mistake” was making substantive changes from the accepted version (which the editor reads) to the version you submit for typesetting (which the editor does not). If the changes were harmless, then why hide them from the editor? I said we are giving this guy the benefit of the doubt… but he is making it hard.
I guess there is a 90% chance this was intentional/malicious data fraud, and 10% chance it was a genuine error.
Let’s be generous and call it a genuine error.
Would he have gotten tenure at Princeton without this “genuine error”?
No, he wouldn’t have.
Now, he gets to keep tenure for the next 50 years.
Just ask Fracesca Gino:
EDIT: After publishing this piece I received the following message from an economist: “I don’t think Adrien Matray has tenure yet.” So I think I was wrong about that tenure.